spong-client - report system information to spong server

spong-client [ --debug|-d n ] [ --kill|--restart|--nosleep|--norefresh ] config-file

This program is run on each Unix machine in which you want to monitor local system attributes, and reports that information to the spong server. It runs one or more configured checks. It then sleeps for a time period you have defined in your configuration file and does it all again (it actually adds or subtracts a random amount of time - no more then 10% of the total you have specified, to keep clients from sync'ing up and overloading the spong-server).

The checks are modular in nature. You can configure the number of checks to run and the order in which to run to them. The list of checks that are included are disk space, cpu load, running processes and log files.

It sends a message for each check to the spong server and reports the following:

* hostname (where is this report coming from)

* a one line summary

* a more detailed message providing additional detail.

The color is determined by comparing the current status of that service against thresholds defined in the configuration file. If they are greater then the level you have defined for a warning, then the color is yellow. If they are greater then the critical level you have defined then the color is red.

The one line summary provides information that might be useful at a glance when looking at the overall system status (such as a brief report on the load, number of users, and uptime).

The more detailed message contains information such as the complete CWdf output, or a listing of the top 10 processes sorted by CPU.

You should start this program in your system startup file, and it should be running constantly. If no parameters are specified, spong-client forks and detaches itself to run as a daemon.

If you provide the --debug n flag, then debugging information will be printed to stdout, otherwise output will only be produced if there is a problem. Where n is a number from 1 - 9. A higher number means more verbosity in the debugging output.

If you provide the --restart flag, a signal will be sent to the spong-client process that is currently running that will cause it to reload it's configuration files. If you provide the --kill flag, a signal will be sent to the running spong-client process causing it to exit.

The --nosleep or --refresh flag causes the program to cycle through all of the checks once then exit. These flags can be used to run spong-client as a cron job (depreciated), this reduces the effectiveness of the check_logs module.

The checks are actually a set of modules that are called in series by spong-client. The list of modules to run are defined in the $CHECKS configuration variable. Upon initialization, spong-client will load the modules defined in $CHECKS from the LIBDIR/Spong/Client/plugins/ directory. As each module is initialized, it registers itself with the the plugins registry (see the Developer Guide).

Depreciated, please refer to CLIENT MODULES in developer-guide.

If you want to check some service which is not being checked by spong-client, then you can define a CW&check_local() function in your config file (either in your standard spong.conf config file or your host specific spong.conf.hostname file - but not both!). That function can do anything you want, but at the end needs to call the CW&status() function to report what you have found to the spong server.

By default this reads the spong.conf file on startup. You can specify an alternate config file via a command line option and it will read that file instead. If you change values in the configuration file you will need to restart this program for those changes to be re-read.

After reading the configuration file that you specify (or the default), it then reads the spong.conf.[hostname] file where [hostname] is the hostname of the machine that you are running on. Since these configuration files are just standard perl code that gets imported, the variables that you define in the host specific config file will take precedence over the standard configuration settings.

Here is a listing of the configuration variables applicable to the spong-client program. Some basic spong configuration options that define how long to sleep (in seconds) before checking the status of a service again, the hostname of the spong server, and the port number that the spong server listens This the new method for specifying the CW$SPONGSLEEP interval for Spong programs. If there is not CW$SPONGSLEEP{} entry for the program, it will use the $SPONGSLEEP{'DEFAULT'} value. If no value is then found, spong-client fall back to using CW$SPONGSLEEP.

$SPONGTMP

The directory that Spong programs use for temporary store and work files. It should be a different directory than /tmp for operation and security reasons.

$SPONG_LOG_FILE

If set to 1, spong-client will log errors to a log file in $SPONGTMP named spong-client.log.

$SPONG_LOG_SYSLOG

If set to 1, spong-client will log errors to the syslog using the USER facility and the ERR priority.

$CHECKS

A string that has the list of client check modules to run. If $CHECKS is missed or blank, spong-client defaults to disk cpu processes logs. If the CWcheck_local() function is present then 'local' is appended. A number indicating the CPU load that triggers a problem ($CPUWARN triggers warnings - yellow, and $CPUCRIT triggers alerts - red). A list of processes that should be running, if they are not running, then trigger a problem (processes in @PROCSWARN trigger a warning - yellow, and processes in @PROCSCRIT trigger an alert - red).

$LOGCHECKS

A list of hashes which defined checks to apply to log files. Each hash contains the fields:

logfile

which is the full path to the log file to check

checks

a list of check to apply to the log file. Each check is a hash that contains the fields:

pattern

a Perl regular expression to be scanned for

status

the status color to reporte when lines match pattern

duration

the duration (in seconds) that each event is to be reported to the server

text

a string which is the the text to be reported back in the detailed message field of the status report (which can include match position variables from pattern)

id

an optional key field to associated with each event generated. The default key is the evaluated text field. An id key may be specified for the for a check pattern. All hits of the pattern will be consolidated into one event. The data of the last hit will be reported in the event. These variables are OS specific variables, which are hopefully set correctly for your machine, if they are not - please send me email letting me know what OS you are running on, and what the correct value should be.

spong.conf, spong.conf.[hostname]

        spong-client --debug 5 --nosleep
        spong-client --debug 5
        spong-client --restart

Perl v5.005_03 or greater is required.

No known bugs.

spong-server, spong.conf, client-modules, developer-guide

Ed Hill <ed-hill@uiowa.edu>, Unix System Administrator, The University of Iowa

Stephen L Johnson <sjohnson@monsters.org>

Based on code/ideas from Sean MacGuire (BB), and Helen Harrison (Pong). Ed Hill original converted Big Brother (http://www.bb4.com) into Perl which diverged from Big Brother to become Spong. Ed Hill continued Spong development until version 2.1. Stephen L Johnson took over development in October, 1999 with his changes which became Spong 2.5.